import { p as EvalMetric, E as EADKAgent, n as EvalCase, R as RunConfig, q as EvalResult } from '../../types-Ks98Z0E_.js';
export { o as EvalContext } from '../../types-Ks98Z0E_.js';
import 'zod';

/**
 * EADK Evaluation Framework
 *
 * Trajectory + response evaluation for agent quality assessment.
 * Google ADK eval pattern.
 */

interface EvalConfig {
    /** Metrics to evaluate */
    metrics: EvalMetric[];
    /** Pass threshold (0-1, default: 0.7) */
    passThreshold?: number;
    /** Run config overrides for eval */
    runConfig?: RunConfig;
    /** Concurrency limit */
    concurrency?: number;
}
interface EvalSuiteResult {
    cases: EvalResult[];
    averageScore: number;
    passRate: number;
    metricAverages: Record<string, number>;
    totalDurationMs: number;
}
/**
 * Evaluate an agent against a set of test cases.
 */
declare function evaluate(agent: EADKAgent, cases: EvalCase[], config: EvalConfig): Promise<EvalSuiteResult>;
/**
 * Exact match metric — checks if output exactly matches expected.
 */
declare const exactMatch: EvalMetric;
/**
 * Contains metric — checks if output contains expected substring.
 */
declare const contains: EvalMetric;
/**
 * Length ratio metric — penalizes outputs that are too short or too long.
 */
declare const lengthRatio: EvalMetric;
/**
 * Tool usage metric — checks if expected tools were called.
 */
declare const toolUsage: EvalMetric;
/**
 * Create a custom regex-based metric.
 */
declare function createRegexMetric(name: string, pattern: RegExp): EvalMetric;

export { EvalCase, type EvalConfig, EvalMetric, EvalResult, type EvalSuiteResult, contains, createRegexMetric, evaluate, exactMatch, lengthRatio, toolUsage };